You are on page 1of 5

Electric field modulation of thermovoltage in

single-layer MoS2
Cite as: Appl. Phys. Lett. 105, 253103 (2014); https://doi.org/10.1063/1.4905014
Submitted: 28 October 2014 • Accepted: 04 December 2014 • Published Online: 23 December 2014

Lukas Dobusch, Marco M. Furchi, Andreas Pospischil, et al.

ARTICLES YOU MAY BE INTERESTED IN

Gate-tunable and thickness-dependent electronic and thermoelectric transport in few-layer


MoS2
Journal of Applied Physics 120, 134305 (2016); https://doi.org/10.1063/1.4963364

Large theoretical thermoelectric power factor of suspended single-layer MoS2


Applied Physics Letters 105, 193901 (2014); https://doi.org/10.1063/1.4901342

Electronic and thermoelectric properties of few-layer transition metal dichalcogenides


The Journal of Chemical Physics 140, 124710 (2014); https://doi.org/10.1063/1.4869142

Appl. Phys. Lett. 105, 253103 (2014); https://doi.org/10.1063/1.4905014 105, 253103

© 2014 Author(s).
APPLIED PHYSICS LETTERS 105, 253103 (2014)

Electric field modulation of thermovoltage in single-layer MoS2


Lukas Dobusch,1 Marco M. Furchi,1 Andreas Pospischil,1 Thomas Mueller,1
Emmerich Bertagnolli,2 and Alois Lugstein2,a)
1
Institute of Photonics, TU-Wien, Gußhausstraße 27-29, A-1040 Vienna, Austria
2
Institute of Solid State Electronics, TU-Wien, Floragasse 7, A-1040 Vienna, Austria
(Received 28 October 2014; accepted 4 December 2014; published online 23 December 2014)
We study electric field modulation of the thermovoltage in single-layer MoS2. The Seebeck
coefficient generally increases for a diminishing free carrier concentration, and in the case of
single-layer MoS2 reaches considerable large values of about S ¼ 5160 lV/K at a resistivity of
490 X m. Further, we observe time dependent degradation of the conductivity in single layer
MoS2, resulting in variations of the Seebeck coefficient. The degradation is attributable to
adsorbates from ambient air, acting as p-dopants and additional Coulomb potentials, resulting in
carrier scattering increase, and thus decrease of the electron mobility. The corresponding power
factors remain at moderate levels, due to the low conductivity of single layer MoS2. However,
as single-layer MoS2 has a short intrinsic phonon mean free path, resulting in low thermal
conductivity, MoS2 holds great promise as high-performance 2D thermoelectric material. V C 2014

Author(s). All article content, except where otherwise noted, is licensed under a Creative
Commons Attribution 3.0 Unported License. [http://dx.doi.org/10.1063/1.4905014]

The discovery and experimental realization of graphene— the thermal conductivity j ¼ je þ jl comprising heat trans-
a single layer of graphite—triggered intense research activities port due to electrons (je) and phonons (jl)
in 2D-material systems.1 A multitude of 2D-materials are
known by now, often showing superior properties compared to rS2 T
zT ¼ : (1)
their bulk counterparts. Due to their direct bandgap, single lay- j
ered transition metal dichalcogenides, such as WS2, WSe2,
Competitive thermoelectric materials must have a zT in
MoSe2, and MoS2, are of high interest for electronic and optoe-
the order of 2, corresponding to an energy conversion of
lectronic applications.2 With respect to MoS2, highly sensitive
approximately 27% of the Carnot limit. The term rS2 is
photodetectors with responsivities up to 880 A/W,3 field effect
referred to as power factor, which, under a given tempera-
transistors (FETs) with an Ion/Ioff ratio of 108,4 flexible tran-
ture, quantifies the ability of a material to produce electrical
sistors,5 and selective gas-sensors6 have been reported. All
power. Semiconductors may exhibit a relatively high zT
these devices take advantage of the peculiar single-layer-
because of their moderate level of free carrier concentra-
related properties of MoS2. With regard to thermoelectric prop-
tions.9 Large carrier concentrations lead to high conductiv-
erties, simulations give reason to presume that a maximum
ities, but generally have a detrimental effect on the Seebeck
Seebeck coefficient of about 1600 lV/K may be achieved in
coefficient, and vice versa for low carrier concentrations.
layered MoS2.7 Recently, Buscema et al.8 suggested that
photocurrent generation in single-layer MoS2 is dominated by Modulation of the carrier concentration and thus tuning of
the photo-thermoelectric effect, indicating remarkably high the thermovoltage of semiconductor materials can be
Seebeck coefficients of up to 1  105 lV/K, obtained by achieved by electrostatic fields, e.g., in FET devices.
laser heating. In order to investigate the electric field modulated thermo-
In this letter, we report on a giant thermovoltage for voltage of single-layer MoS2, these were integrated into back-
single-layer MoS2, which can be tuned by an external elec- gated FET devices, equipped with resistive heating structures
tric field. By doing so, a thermal bias along single-layer as shown in the scanning electron micrograph in Figure 1(a).
MoS2 integrated in a back gated FET structure is set up by In addition, this structure enables four-point measurements on
Joule heating and measured in situ by two resistive thermom- the MoS2 monolayer allowing precise in situ measurements of
eters, allowing precise determination of the Seebeck coeffi- the electric field modulated MoS2 resistivity.
cient. A maximal Seebeck coefficient of 5160 lV/K is For device fabrication, single-layer MoS2 flakes were
thus demonstrated, making MoS2 a promising material for deposited by mechanical exfoliation onto highly doped Si
thermoelectric applications. substrates with a 300 nm thick SiO2 layer. Characterisation
The effectiveness of thermoelectric materials can be of the MoS2 flakes, typically 10 lm2–50 lm2 in size, was
linked to the dimensionless figure of merit zT, which is pro- done by optical microscopy10–12 and Raman spectros-
portional to the Seebeck coefficient S, electrical conductivity copy.13,14 The electrical contacts and resistive heater struc-
r, and absolute temperature T, but inversely proportional to tures were fabricated by electron beam lithography, sputter
metal deposition of Ti/Au (5 nm/130 nm) and lift-off techni-
ques. Prior to the thermo-physical characterization, all
a)
Author to whom correspondence should be addressed. Electronic mail: samples were annealed at 200  C for 2 h in He/H2 flow (100
alois.lugstein@tuwien.ac.at sccm/20 sccm) in order to improve contact reliability.4

0003-6951/2014/105(25)/253103/4 105, 253103-1 C Author(s) 2014


V
253103-2 Dobusch et al. Appl. Phys. Lett. 105, 253103 (2014)

heater structures were operated at a heating power of approxi-


mately 180 mW leading to thermal biases between 2 K and
6 K, depending on the respective device geometry. The tem-
perature dependency of the resistive thermometers was deter-
mined in separate four-point probe measurements.14 The
temperature coefficient was determined to DR  0.03 X/K
 DT, enabling temperature measurements with an accuracy
of 60.34 K.
Figure 2 shows typical I/V measurements, with a ther-
mal bias of DT ¼ 2.3 K (air, 296 K). Sweeping the backgate
voltage (VBG) in alternating directions, a full I/V sweep was
recorded at each gating value.
Figure 2 and the enlarged view in the upper inset shows
a gate-dependent shift of the I/V curve when a thermal bias
is applied. As expected in the reference measurement, under
isothermal conditions (DT ¼ 0 K), shown in the lower inset
of Figure 2 no such shift of the I/V curve was observed inde-
pendent from the back gate voltage. This shift along the
VDS-axis can be attributed to the thermovoltage VTE of
the MoS2-Au thermocouple. As expected, by reversing the
temperature gradient, a shift in the opposite direction was
obtained.14 The small contribution of the Au-electrodes to
the thermovoltage, which is independent of the gate voltage,
can be subtracted by using

VTE ¼ ðSMoS2  SAu ÞDT; (2)

with SAu ¼ T  6:4 nV=K2 .17,18


To investigate the influence of electric field modulation,
a backgate sweep was performed and the induced conductiv-
FIG. 1. (a) SEM image of a typical thermoelectric test device. The single- ity modulation of the single-layer MoS2 flake was evaluated
layer MoS2 is connected by resistive thermometers (Ti/Au), acting also as using
drain and source contacts of the backgated FET device, as well as the four-
point measurement module. A temperature gradient along the MoS2 flake is l
achieved by resistive heater elements, inducing a thermovoltage between r ¼ ð R  2Rc Þ1 ; (3)
source and drain. The thermal bias DT is determined via the resistive Ti/Au wd
thermometers in a four-point measurement configuration. (b) Electrical per-
formance of a single-layer MoS2 FET with a flake of length l ¼ 6 lm and a where R and Rc are the device and contact resistance, respec-
width of w ¼ 5.5 lm. IDS/VDS curves are shown as a function of the tively, d denotes the thickness of single-layer MoS2 of
backgate-voltage and the respective transfer response (IDS/VBG) for a full
backgate sweep from VBG ¼ 100 to VBG ¼ 100 V is shown in the inset.

Typical output and transfer characteristics are shown in


Figure 1(b) and the inset, respectively. The back-gated sin-
gle-layer MoS2 FET exhibits n-type behaviour with excellent
field modulation response and a remarkable high ION/IOFF
ratio of 106, in good agreement with previous reports.4,8 The
output characteristic indicates nonlinear behaviour, which
might be attributable to space-charge-limited current15 or
Schottky type16 behaviour of the MoS2-Au contacts. The re-
sistance of these semiconductor-metal contacts, investigated
by four-point measurements, show a strong dependence on
the backgate voltage.14 This field modulated response of the
metallic contacts is attributed to band-bending, its accompa-
nied change in tunnel-efficiency and electrical doping of the
MoS2-Layer, and was taken into account for all succeeding
thermo-physical measurements.
To determine the Seebeck coefficient and to investigate FIG. 2. I/V characteristics of a backgated MoS2 device with a thermal bias
the field modulation response, the output characteristics of the of DT ¼ 2.3 K as a function of VBG, varied from 40 V to 80 V in steps of
5 V. The intercept at IDS¼0 A allows to extract the thermovoltage VTE. The
backgated MoS2-FET devices were recorded at low VDS with
enlarged view in the upper inset shows the gate dependency of the thermo-
and without a thermal bias between drain and source. voltage in more detail (linear fit). The lower inset shows the same device
Differential heating was obtained when the resistive Joule under isothermal conditions (DT ¼ 0 K) exhibiting no current/voltage shift.
253103-3 Dobusch et al. Appl. Phys. Lett. 105, 253103 (2014)

0.65 nm,4 and l and w are the channel length and width, degradation effects, or defects within the MoS2 layer result-
respectively, which were determined by subsequent SEM ing in time dependent variations of the thermoelectric
imaging. As shown in the upper inset of Figure 2, the ther- properties.9 In order to explore these degradation effects, we
movoltage increases with decreasing backgate voltage VBG. measured the transfer characteristic for an as processed
Figure 3 shows thereof calculated Seebeck coefficients and sample as well as for the same sample after several days of
their corresponding power factors as a function of resistivity, storage in ambient air. Figure 4 shows that the transfer char-
modulated by the back gate sweep. acteristics of the field modulated single-layer MoS2 device
With decreasing backgate voltage, the resistivity of the exhibits a significant hysteresis in the VBG dependence as a
n-type single-layer MoS2 device increases (see Figure 1(b)) consequence of the slow dynamics of surface states and
and thus the Seebeck coefficient. The Seebeck coefficient traps, leading to a time dependent screening of the gate.
generally increases for a diminishing free carrier concentra- However, for the latter measurement one clearly sees a
tion,9,19,20 and in case of single-layer MoS2 reaches particu- decrease of the on-current by one order of magnitude, broad-
larly large values (>500 lV/K) as the sample resistivity ening of the hysteresis and a pronounced shift of the thresh-
increases above 50 X m. A considerable high Seebeck old voltage towards positive VBG.
coefficient was observed at a resistivity of 490 X m with The time resolved resistivity measurement, plotted in
S ¼ 5160 lV/K. As shown in the inset of Figure 3, the rise the inset of Figure 4, demonstrates a continuous increase of
of the thermovoltage was observed for all investigated sam- the resistivity by almost one order of magnitude within the
ples, which evidently gives the possibility of tuning the first 44 h after sample processing and the terminal thermal
Seebeck coefficient by field induced resistivity modulation. annealing. Note that this measurement was performed in a
Even though the Seebeck coefficients at negative back- dark box, and the time dependent resistivity measurements
gate voltage, i.e., the high resistive regime, exhibit exceed- were started 30 min after closing the darkbox in order to
ingly large values, the corresponding power factors remain exclude photoconductivity effects.14,22
moderate. For positive backgate voltage (i.e., in the low In accordance with Zhang et al.,22 we suppose that the
resistivity regime), the power factor shows values of a few observed degradation effect is due to adsorbates from ambient
100 nW/K2 m (see Figure 3), whereas the power factors at air which interact strongly with MoS2, resulting in enhanced
negative backgate voltage are about one order of magnitude carrier scattering and reduced mobility. By introducing the de-
lower. However, calculations by Liu et al. show that single- vice in vacuum and thus removing the adsorbates, the on state
layer MoS2 has a very short intrinsic phonon mean free path, current increases by one order of magnitude, and the threshold
resulting in a remarkably low thermal conductivity of voltage of the FET is shifted towards negative backgate
1.35 W/mK.21 Thus, despite the low power factor, MoS2 bias.14 Thus, adsorbates on the surface or at the MoS2/sub-
holds great promise as high-performance 2D thermoelectric strate interface appear to act as p-dopants for MoS2 and addi-
tional Coulomb potentials resulting in carrier scattering
material.
increase, thus lowering electron mobility. Recovery of the
The plot of the field modulated thermovoltage, shown in
device performance can also be achieved by annealing of the
the inset of Figure 3, illustrates distinct variations for each
device in H2/He atmosphere at 200  C for 120 min.
individual single-layer MoS2 device. Further, time dependent
measurements revealed a continuous increase of the Seebeck
coefficient for a given backgate voltage within the first hours
after device processing and the finalizing annealing proce-
dure. We suppose that these variations are the result of

FIG. 4. Transfer characteristic of a backgated single-laver MoS2 device


FIG. 3. Semilog plot demonstrating the field modulation of the Seebeck straight after processing and annealing (black solid curve), respectively, and
coefficient (black squares) and the corresponding power factor (red circles) after storage of the sample for 20 days in ambient atmosphere (red dashed
for a typical single-layer MoS2 device. The Seebeck coefficient varies curve). (VDS ¼ 2 V, sweeping the backgate voltage forward and reverse as
between 118 lV/K and 5160 lV/K as a function of the resistivity modu- indicated by the arrows.) The time dependent resistivity measurement in the
lated by sweeping the back gate voltage. The inset shows determined inset was recorded via a four-point measurement method (VDS ¼ 2 V, VBG
Seebeck coefficients for 8 investigated single-layer MoS2 thermoelectric ¼ 0 V) straight after annealing and shows a long time degradation
devices. behaviour.
253103-4 Dobusch et al. Appl. Phys. Lett. 105, 253103 (2014)

5
In conclusion, we demonstrate a giant thermovoltage for G.-H. Lee, Y.-J. Ju, X. Cui, N. Petrone, C.-H. Lee, M. S. Choi, D.-Y. Lee,
C. Lee, W. J. Yoo, K. Watanabe et al., ACS Nano 7, 7931 (2013).
single-layer MoS2, which can be tuned by an external electric 6
F.-K. Perkins, A. L. Friedman, E. Cobas, P. M. Campbell, G. G. Jernigan,
field with a maximum Seebeck coefficient of 5160 lV/K, and B. T. Jonker, Nano Lett. 13, 668 (2013).
7
making MoS2 a promising material for thermoelectric appli- C. Lee, J. Hong, M. H. Whangbo, and J. H. Shim, Chem. Mater. 25, 3745
cations. The corresponding power factors remain at moderate (2013).
8
M. Buscema, M. Barkelid, V. Zwiller, H. S. J. van der Zant, G. A. Steele,
levels, due to the low conductivity of single layer MoS2. and A. Castellanos-Gomez, Nano Lett. 13, 358 (2013).
However, the electric conductivity, and hence the power fac- 9
G. J. Snyder and E. S. Toberer, Nat. Mater. 7, 105 (2008).
10
tor, might be significantly increased by a high-k dielectric M. M. Benameur, B. Radisavljevic, J. S. Heron, S. Sahoo, H. Berger, and
passivation as shown in recent works.4,23 Together with the 11
A. Kis, Nanotechnology 22, 125706 (2011).
H. Li, J. Wu, X. Huang, G. Lu, J. Yang, X. Lu, Q. Xiong, and H. Zhang,
low thermal conductivity of single-layer MoS2,21 zT might ACS Nano 7, 10344 (2013).
therefore show superior high values. 12
A. Castellanos-Gomez, N. Agra€ıt, and G. Rubio-Bollinger, Appl. Phys.
The mean variation of the thermovoltage was assigned Lett. 96, 213116 (2010).
13
C. Lee, H. Yan, L. E. Brus, T. F. Heinz, J. Hone, and S. Ryu, ACS Nano 4,
to adsorbates, interacting strongly with MoS2. These adsor-
2695 (2010).
bates act as p-dopants and thus additional Coulomb poten- 14
See supplementary material at http://dx.doi.org/10.1063/1.4905014 for
tials, resulting in carrier scattering increase, thus lowering materials and methods employed, such as Joule heater operation, method
the electron mobility. of thermovoltage measurement, contact resistance, photoconductivity and
degradation effects.
15
S. Ghatak and A. Ghosh, Appl. Phys. Lett. 103, 122103 (2013).
The authors acknowledge financial support by the 16
H. Liu, A. T. Neal, and P. D. Ye, ACS Nano 6, 8563 (2012).
17
Austrian Science Fund (FWF): project No. I 724-N16. The S. Roddaro, D. Ercolani, M. A. Safeen, S. Suomalainen, F. Rossella, F.
authors further thank the Center for Micro- and Nanostructures Giazotta, L. Sorba, and F. Beltram, Nano Lett. 13, 3638 (2013).
18
N. F. Mott and H. Jones, The Theory of the Properties of Metals and
(ZMNS) for providing the clean-room facilities. Alloys (Dover Publication Inc., 1936).
19
W. Liang, A. I. Hochbaum, M. Fardy, O. Rabin, M. Zhang, and P. Yang,
1
K. S. Novoselov, D. Jiang, F. Schedin, T. J. Booth, V. V. Khotkevich, Nano Lett. 9, 1689 (2009).
20
S. V. Morozov, and A. K. Geim, Proc. Natl. Acad. Sci. U.S.A. 102, 10451 P. M. Wu, J. Gooth, X. Zianni, S. F. Svensson, J. G. Gluschke, K. A.
(2005). Dick, C. Thelander, K. Nielsch, and H. Linke, Nano Lett. 13, 4080
2
Q. H. Wang, K. Kalantar-Zadeh, A. Kis, J. N. Coleman, and M. S. Strano, (2013).
21
Nat. Nanotechnol. 7, 699 (2012). X. Liu, G. Zhang, Q.-X. Pei, and Y.-W. Zhang, Appl. Phys. Lett. 103,
3
O. Lopez-Sanchez, D. Lembke, M. Kayci, A. Radenovic, and A. Kis, Nat. 133113 (2013).
22
Nanotechnol. 8, 497 (2013). W. Zhang, J.-K. Huang, C.-H. Chen, Y.-H. Chang, Y.-J. Cheng, and L.-J.
4
B. Radisavljevic, A. Radenovic, J. Brivio, V. Giacometti, and A. Kis, Nat. Li, Adv. Mater. 25, 3456 (2013).
23
Nanotechnol. 6, 147 (2011). B. Radisavljevic and A. Kis, Nat. Mater. 12, 815 (2013).

You might also like